Está en la página 1de 36

Model building with Python–

Regression analysis

Sushrutha Raj
PhD Integrative Sciences, Amity University Gurgaon | June 2018
Outline

1. Understanding Regression
2. Definition
3. Objective
4. Univariate or Simple linear regression
1. Understanding SLS
2. Model Assumptions
3. Case Study
5. Multivariate linear regression
1. Understanding SLS
2. Model Assumptions
3. Case Study
Regression analysis
Understanding Regression analysis

 Linear regression analysis means“ fitting a straight line to data” , also called
Linear Modelling

 It’s a widely used technique to help model data and understand real-world
phenomena
a) Easy to Use (Build)
b) Easy to interpret

 Allows prediction
Definition

 A regression problem is composed of


• an outcome or response variable Y
• a number of risk factors or predictor variables Xi that affect Y ,
also called explanatory variables or features

Y is sometimes called the dependent variable and Xi the


independent variables

For instance:
1. Yield depends on like Rainfall, Fertilizers, Soil fertility etc
2. IQ of a student on School, Teachers, Environment etc
Objectives
Univariate linear regression
(when all you have is a single predictor variable)
Linear regression analysis

 A regression problem is composed of


• an outcome or response variable Y
• a number of risk factors or predictor variables Xi that affect Y ,
also called explanatory variables or features

 Y is sometimes called the dependent variable and Xi the


independent variables

 For instance:
1. Yield depends on like Rainfall, Fertilizers, Soil fertility etc
2. IQ of a student on School, Teachers, Environment etc
Linear regression analysis

Makes strong assumptions


about the relationship
between the predictor
variables (Xi ) and the
response (Y )
Linear regression:
one of the simplest a linear relationship, a
and most commonly straight line when
plotted
used statistical
modeling techniques only valid for
continuous outcome
variables (not
applicable to category
outcomes
Linear regression analysis

 Assumption: y = 𝛽0 +𝛽1 ×x +error

 Our task: estimate 𝛽0 and 𝛽1 based on the available data

 Resulting model is ŷ = 𝛽0 ̂ + 𝛽1 ̂* x
 the “hats” on the variables represent the fact that they are estimated from the available data
 ŷ is read as “the estimator for y”

 𝛽0 and 𝛽1 are called the model parameters or


coefficients

 Objective: minimize the error, the difference between


our observations and the predictions made by our.
linear model
 Minimize the length of the red lines in the figure to the right.
(called the “residuals”)
Case Study 1:
Simple linear regression analysis –
Data description
Simple linear regression analysis
Simple linear regression analysis –
Model Building
Simple linear regression analysis –
Model Building
Case Study 2:
Simple linear regression analysis

British Doctors’ Study followed the health of a large number


of physicians in the UK over the period 1951–2001
Provided conclusive evidence of linkage between smoking
and lung cancer, myocardial infarction, respiratory disease
and other illnesses
Provides data on annual mortality for a variety of diseases at
four levels of cigarette smoking:
1. never smoked
2. 1-14 per day
3. 15-24 per day
4. 25perday
Simple linear regression analysis –
Data description

Cigarettes Lung cancer


smoked CVD mortality mortality
(per 100 000 men per (per 100 000 men per
(per day) year) year)
0 572 14
10 (actually 1-14) 802 105

20 (actually 15-24) 892 208


30 (actually >24 1025 355
Simple linear regression analysis
Simple linear regression analysis

In 1964: the US Surgeon General issues a


report claiming that cigarette smoking
causes lung cancer, based mostly on
correlation data similar to the previous slide.
Smoking
causes Lung
cancer !!!!!!

https://profiles.nlm.nih.gov/ps/retrieve/Narrative/NN/p-nid/60
Simple linear regression analysis

1964: the US Surgeon General issues a


report claiming that cigarette smoking
causes lung cancer, based mostly on
correlation data similar to the previous
slide.

However, correlation is not sufficient to


demonstrate causality. ere might be some
hidden genetic factor that causes both lung
cancer and desire for nicotine.
Simple linear regression analysis –
Model Building
Simple linear regression analysis –
Prediction

Q: What is the expected lung cancer mortality risk for a group of people who smoke 15
cigarettes per day?
Simple linear regression analysis -
Summary
Multiple linear regression
(when all you have is a multiple predictor variables)
Understanding Multiple linear
regression

Multivariate linear regression involves more than one predictor variables


For Instance, Yield of a crop not just depends on quality of soil
Multiple linear regression- Model

 WKT the equation for univariate linear regression: ŷ = 𝛽0 + 𝛽1x

 Equation for multivariate linear regression:


ŷ = 𝛽0 +𝛽1x1 +𝛽2x2 +...+𝛽nxn

 The outcome variable is assumed to be a linear combination of the predictor


variables (the inputs)
Case Study
Multiple linear regression- Data
description

 Data set contains 9568 data points collected from a combined cycle power plant over 6
years, when power plant was under full load
 A combined cycle power plant is composed of gas turbines, steam turbines and heat
recovery steam generators
 Electricity is generated by gas & steam turbines, which are combined in one cycle
 Three ambient variables affect performance of the gas turbine
 Exhaust vacuum affects performance of the steam turbine
 Data consists of hourly averages taken from various sensors located around the plant that
record the ambient variables every second
 Let’s load it into Python and examine it using the pandas library

http://archive.ics.uci.edu/ml/machine-learning-databases/00294/
Multiple linear regression- Data
description

 Objective: predict energy output at a Combined Cycle Power Plant


 Data available: hourly averages of variables

 Let’s try to build a multivariate linear model to predict PE given


inputs AT, AP, RH and V

http://archive.ics.uci.edu/ml/machine-learning-databases/00294/
Multiple linear regression- Data
description

http://archive.ics.uci.edu/ml/machine-learning-databases/00294/
Multiple linear regression- Data
description

http://archive.ics.uci.edu/ml/machine-learning-databases/00294/
Multiple linear regression- Data
description

http://archive.ics.uci.edu/ml/machine-learning-databases/00294/
Multiple linear regression – Model
Building
Multiple linear regression –
Prediction and Summary